摘要
针对传统主动轮廓模型难以实现煤矿井下早期火灾图像火焰区域精确提取的问题,提出了一种改进的Chan-Vese(CV)模型.在计算目标和背景全局区域拟合中心的基础上,利用曲线内外区域局部灰度统计直方图获取目标和背景局部区域拟合中心,并对全局和局部区域拟合中心赋予归一化的调节比例,以综合利用图像全局和局部信息;为了加速曲线运动到目标边缘,利用曲线内外区域像素灰度的最小绝对差来取代模型中原有的内外区域能量权重,以提高模型分割效率.结果表明:与CV模型、局部二值拟合模型(LBF)、全局和局部灰度拟合混合模型(LGIF)、引入自适应能量权重的CV模型(WCV)相比较,提出的模型能更加快速、精确地提取煤矿井下早期火灾图像中的火焰区域,在分割效果和分割效率方面均有明显优势.
It is difficult to utilize traditional active contour models to extract the fire region of the early fire image of mine accurately.Aiming at this problem,an improved CV model was proposed.On the basis of calculating the global region fitting centers of the object and background regions,the local gray histograms of the region inside and outside the curve were utilized to obtain the local region fitting centers of the object and background regions.Then the normalized adjustable ratios were incorporated into the global and local region fitting centers,which can synthetically utilize both global and local information of the image.To accelerate the motion of the curve towards the object boundaries,the minimum absolute differences of the pixel grayscale values inside and outside the curve were used to replace the original internal and external energy weights,which can improve the segmentation efficiency of the model.The experimental results show that,compared with CV model,LBF model,LGIF model,and the CV model incorporating adaptive energy weight,the proposed model can extract the fire region of the early fire image of mine more rapidly and accurately,and has obvious advantages in terms of segmentation performance and segmentation efficiency.
出处
《中国矿业大学学报》
EI
CAS
CSCD
北大核心
2018年第2期429-435,共7页
Journal of China University of Mining & Technology
基金
国家自然科学基金项目(61573183)
煤矿安全高效开采省部共建教育部重点实验室开放基金项目(JYBSYS2014102)
关键词
矿井
早期火灾图像
图像分割
主动轮廓模型
区域拟合中心
最小绝对差
mine
early fire image
image segmentation
active contour model
region fitting center
minimum absolute difference
